platform/adaptation/renesas_rcar/renesas_kernel.git
10 years agof2fs: reorganise the function get_victim_by_default
Namjae Jeon [Sat, 1 Jun 2013 07:20:26 +0000 (16:20 +0900)]
f2fs: reorganise the function get_victim_by_default

Fix the function get_victim_by_default, where it checks
for the condition  that p.min_segno != NULL_SEGNO as
shown:

if (p.min_segno != NULL_SEGNO)
           goto got_it;

and if above condition is true then

got_it:
        if (p.min_segno != NULL_SEGNO) {

So this condition is being checked twice. Hence move the goto
statement after the if condition so that duplication of condition
check is avoided.

Also this function makes a call to get_max_cost() to compute
the max cost based on the f2fs_sbi_info and victim policy. Since
get_max_cost depends on on three parameters of victim_sel_policy
=> alloc_mode, gc_mode & ofs_unit, once this victim policy is
initialised, these value will not change till the execution
time of get_victim_by_default() & also f2fs_sbi_info structure
parameters will not change.

Hence making calls to get_max_cost() in while loop does not seems to
be a good point. Instead we can call it once in begining and store
the results in local variable, which later can serve our purpose
for comparing the cost with max cost inside the while loop.

Signed-off-by: Namjae Jeon <namjae.jeon@samsung.com>
Signed-off-by: Pankaj Kumar <pankaj.km@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
10 years agof2fs: handle errors from get_node_page calls
Jason Hrycay [Fri, 31 May 2013 17:45:11 +0000 (12:45 -0500)]
f2fs: handle errors from get_node_page calls

Add check for error pointers returned from get_node_page in order to
avoid dereferencing a bad address on the next use.

Signed-off-by: Jason Hrycay <jason.hrycay@motorola.com>
Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
11 years agof2fs: cover cp_file information with ilock
Jaegeuk Kim [Tue, 28 May 2013 03:25:47 +0000 (12:25 +0900)]
f2fs: cover cp_file information with ilock

If a file is linked with other files, it should be checkpointed at every fsync
calls.
For this, we use set_cp_file() with FADVISE_CP_BIT, but previously we didn't
cover the flag by the global lock.
This patch fixes that the inode page stores this correctly.

Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
11 years agof2fs: fix incorrect iputs during the dentry recovery
Jaegeuk Kim [Tue, 28 May 2013 00:59:27 +0000 (09:59 +0900)]
f2fs: fix incorrect iputs during the dentry recovery

- iget/iput flow in the dentry recovery process

1. *dir* = f2fs_iget
2. set FI_DELAY_IPUT to *dir*
3. add *dir* to the dirty_dir_list
   - __f2fs_add_link
     - recover_dentry)
4. iput *dir* by remove_dirty_dir_inode
   - sync_dirty_dir_inodes
     - write_chekcpoint

If *dir*'s i_count is not 1 (i.e., root dir), remove_dirty_dir_inode is called
later and then iput is triggered again due to the FI_DELAY_IPUT flag.
So, let's unset the flag properly once iput is triggered.

Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
11 years agof2fs: fix dentry recovery routine
Jaegeuk Kim [Tue, 28 May 2013 00:19:22 +0000 (09:19 +0900)]
f2fs: fix dentry recovery routine

The error scenario is:
1. create /a
(1.a link /a /b)
2. sync
3. unlinke /a
4. create /a
5. fsync /a
6. Sudden power-off

When the f2fs recovers the fsynced dentry, /a, we discover an exsiting dentry at
f2fs_find_entry() in recover_dentry().

In such the case, we should unlink the existing dentry and its inode
and then recover newly created dentry.

Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
11 years agof2fs: iput only if whole data blocks are flushed
Jaegeuk Kim [Mon, 27 May 2013 01:32:01 +0000 (10:32 +0900)]
f2fs: iput only if whole data blocks are flushed

If there remains some unwritten blocks from the recovery, we should not call
iput on that directory inode.
Otherwise, we can loose some dentry blocks after the recovery.

Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
11 years agof2fs: return proper error from start_gc_thread
Namjae Jeon [Sun, 26 May 2013 02:05:32 +0000 (11:05 +0900)]
f2fs: return proper error from start_gc_thread

when there is an error from kthread_run, then return proper error
rather than returning -ENOMEM.

Signed-off-by: Namjae Jeon <namjae.jeon@samsung.com>
Signed-off-by: Amit Sahrawat <a.sahrawat@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
11 years agof2fs: optimize several routines in node.h
Namjae Jeon [Thu, 23 May 2013 13:58:40 +0000 (22:58 +0900)]
f2fs: optimize several routines in node.h

There are various functions with common code which could be separated
out to make common routines. So, made new routines and in order to
retain the same call path and no major changes, written some macros
to access those routines.

Signed-off-by: Namjae Jeon <namjae.jeon@samsung.com>
Signed-off-by: Amit Sahrawat <a.sahrawat@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
11 years agof2fs: remove unneeded initializations in f2fs_parent_dir
Namjae Jeon [Thu, 23 May 2013 13:58:07 +0000 (22:58 +0900)]
f2fs: remove unneeded initializations in f2fs_parent_dir

There is no need to initialize few pointers in f2fs_parent_dir
as the values are not checked and instead directly initialized
values are used.

Signed-off-by: Namjae Jeon <namjae.jeon@samsung.com>
Signed-off-by: Amit Sahrawat <a.sahrawat@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
11 years agof2fs: push some variables to debug part
Namjae Jeon [Thu, 23 May 2013 13:57:53 +0000 (22:57 +0900)]
f2fs: push some variables to debug part

Some, counters are needed only for the statistical information
while debugging.
So, those can be controlled using CONFIG_F2FS_STAT_FS,
pushing the usage for few variables under this flag.

Signed-off-by: Namjae Jeon <namjae.jeon@samsung.com>
Signed-off-by: Amit Sahrawat <a.sahrawat@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
11 years agof2fs: align data types between on-disk and in-memory block addresses
Jaegeuk Kim [Fri, 24 May 2013 03:41:04 +0000 (12:41 +0900)]
f2fs: align data types between on-disk and in-memory block addresses

The on-disk block address is defined as __le32, but in-memory block address,
block_t, does as u64.

Let's synchronize them to 32 bits.

Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
11 years agof2fs: dereferencing an ERR_PTR
Dan Carpenter [Thu, 23 May 2013 10:02:13 +0000 (13:02 +0300)]
f2fs: dereferencing an ERR_PTR

There is an error path where "dir" is an ERR_PTR.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
11 years agof2fs: use ihold
Jaegeuk Kim [Wed, 22 May 2013 03:06:26 +0000 (12:06 +0900)]
f2fs: use ihold

Use the following helper function committed by Al.

commit 7de9c6ee3ecffd99e1628e81a5ea5468f7581a1f
Author: Al Viro <viro@zeniv.linux.org.uk>
Date:   Sat Oct 23 11:11:40 2010 -0400

    new helper: ihold()

...

Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
11 years agof2fs: should not make_bad_inode on f2fs_link failure
Jaegeuk Kim [Wed, 22 May 2013 03:03:47 +0000 (12:03 +0900)]
f2fs: should not make_bad_inode on f2fs_link failure

If -ENOSPC is met during f2fs_link, we should not make the inode as bad.
The inode is still alive.

Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
11 years agof2fs: fix to handle do_recover_data errors
Jaegeuk Kim [Tue, 21 May 2013 23:20:01 +0000 (08:20 +0900)]
f2fs: fix to handle do_recover_data errors

This patch adds error handling codes of check_index_in_prev_nodes and its
caller, do_recover_data.

Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
11 years agof2fs: reuse the locked dnode page and its inode
Jaegeuk Kim [Tue, 21 May 2013 23:02:02 +0000 (08:02 +0900)]
f2fs: reuse the locked dnode page and its inode

This patch fixes the following deadlock bug during the recovery.

INFO: task mount:1322 blocked for more than 120 seconds.
"echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
mount           D ffffffff81125870     0  1322   1266 0x00000000
 ffff8801207e39d8 0000000000000046 ffff88012ab1dee0 0000000000000046
 ffff8801207e3a08 ffff880115903f40 ffff8801207e3fd8 ffff8801207e3fd8
 ffff8801207e3fd8 ffff880115903f40 ffff8801207e39d8 ffff88012fc94520
Call Trace:
[<ffffffff81125870>] ? __lock_page+0x70/0x70
[<ffffffff816a92d9>] schedule+0x29/0x70
[<ffffffff816a93af>] io_schedule+0x8f/0xd0
[<ffffffff8112587e>] sleep_on_page+0xe/0x20
[<ffffffff816a649a>] __wait_on_bit_lock+0x5a/0xc0
[<ffffffff81125867>] __lock_page+0x67/0x70
[<ffffffff8106c7b0>] ? autoremove_wake_function+0x40/0x40
[<ffffffff81126857>] find_lock_page+0x67/0x80
[<ffffffff8112698f>] find_or_create_page+0x3f/0xb0
[<ffffffffa03901a8>] ? sync_inode_page+0xa8/0xd0 [f2fs]
[<ffffffffa038fdf7>] get_node_page+0x67/0x180 [f2fs]
[<ffffffffa039818b>] recover_fsync_data+0xacb/0xff0 [f2fs]
[<ffffffff816aaa1e>] ? _raw_spin_unlock+0x3e/0x40
[<ffffffffa0389634>] f2fs_fill_super+0x7d4/0x850 [f2fs]
[<ffffffff81184cf9>] mount_bdev+0x1c9/0x210
[<ffffffffa0388e60>] ? validate_superblock+0x180/0x180 [f2fs]
[<ffffffffa0387635>] f2fs_mount+0x15/0x20 [f2fs]
[<ffffffff81185a13>] mount_fs+0x43/0x1b0
[<ffffffff81145ba0>] ? __alloc_percpu+0x10/0x20
[<ffffffff811a0796>] vfs_kern_mount+0x76/0x120
[<ffffffff811a2cb7>] do_mount+0x237/0xa10
[<ffffffff81140b9b>] ? strndup_user+0x5b/0x80
[<ffffffff811a3520>] SyS_mount+0x90/0xe0
[<ffffffff816b3502>] system_call_fastpath+0x16/0x1b

The bug is triggered when check_index_in_prev_nodes tries to get the direct
node page by calling get_node_page.
At this point, if the direct node page is already locked by get_dnode_of_data,
its caller, we got a deadlock condition.

This patch adds additional condition check for the reuse of locked direct node
pages prior to the get_node_page call.

Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
11 years agof2fs: fix wrong condition check
Jaegeuk Kim [Tue, 21 May 2013 01:17:56 +0000 (10:17 +0900)]
f2fs: fix wrong condition check

While an orphan inode has zero link_count, f2fs_gc is able to select the inode
for foreground gc.

- f2fs_gc
 - do_garbage_collect
   - gc_data_segment
     : f2fs_iget is failed
     : get_valid_blocks() != 0, so that retry
--> here we got the infinite loop.

This patch resolved this issue.

Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
11 years agof2fs: add f2fs_readonly()
Jaegeuk Kim [Mon, 20 May 2013 11:28:47 +0000 (20:28 +0900)]
f2fs: add f2fs_readonly()

Introduce a simple macro function for readability.

Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
11 years agof2fs: avoid RECLAIM_FS-ON-W: deadlock
Jaegeuk Kim [Mon, 20 May 2013 07:15:22 +0000 (16:15 +0900)]
f2fs: avoid RECLAIM_FS-ON-W: deadlock

This patch tries to avoid the following deadlock condition of which the reclaim
path can trigger f2fs_balance_fs again.

=================================
[ INFO: inconsistent lock state ]
---------------------------------
inconsistent {RECLAIM_FS-ON-W} -> {IN-RECLAIM_FS-W} usage.
kswapd0/41 [HC0[0]:SC0[0]:HE1:SE1] takes:
 (&sbi->gc_mutex){+.+.?.}, at: f2fs_balance_fs+0xe6/0x100 [f2fs]
{RECLAIM_FS-ON-W} state was registered at:
  [<ffffffff810aa5a9>] mark_held_locks+0xb9/0x140
  [<ffffffff810aae85>] lockdep_trace_alloc+0x85/0xf0
  [<ffffffff8113ab2c>] __alloc_pages_nodemask+0x7c/0x9b0
  [<ffffffff81175aa8>] alloc_pages_current+0xb8/0x180
  [<ffffffff811319cf>] __page_cache_alloc+0xaf/0xd0
  [<ffffffff8113225c>] find_or_create_page+0x4c/0xb0
  [<ffffffffa021359e>] find_data_page+0x14e/0x210 [f2fs]
  [<ffffffffa021161b>] f2fs_gc+0x9eb/0xd90 [f2fs]
  [<ffffffffa0218fae>] f2fs_balance_fs+0xee/0x100 [f2fs]
  [<ffffffffa020848c>] f2fs_setattr+0x6c/0x200 [f2fs]
  [<ffffffff811ae51b>] notify_change+0x1db/0x3a0
  [<ffffffff8118fbd0>] do_truncate+0x60/0xa0
  [<ffffffff8118fd95>] vfs_truncate+0x185/0x1b0
  [<ffffffff8118fe1c>] do_sys_truncate+0x5c/0xa0
  [<ffffffff8118ffee>] SyS_truncate+0xe/0x10
  [<ffffffff816e2b42>] system_call_fastpath+0x16/0x1b

Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
11 years agof2fs: don't do checkpoint if error is occurred
Jaegeuk Kim [Mon, 20 May 2013 05:48:49 +0000 (14:48 +0900)]
f2fs: don't do checkpoint if error is occurred

If we met an error during the dentry recovery, we should not conduct checkpoint.
Otherwise, some errorneous dentry blocks overwrites the existing blocks that
contain the remaining recovery information.

Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
11 years agof2fs: fix to unlock page before exit
Jaegeuk Kim [Mon, 20 May 2013 01:26:09 +0000 (10:26 +0900)]
f2fs: fix to unlock page before exit

If we got an error after lock_page, we should unlock it before exit.

Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
11 years agof2fs: remove unnecessary kmap/kunmap operations
Jaegeuk Kim [Mon, 20 May 2013 01:23:40 +0000 (10:23 +0900)]
f2fs: remove unnecessary kmap/kunmap operations

The allocated page used by the recovery is not on HIGHMEM, so that we don't
need to use kmap/kunmap.

Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
11 years agof2fs: reorganize f2fs_vm_page_mkwrite
Namjae Jeon [Sun, 28 Apr 2013 00:04:18 +0000 (09:04 +0900)]
f2fs: reorganize f2fs_vm_page_mkwrite

Few things can be changed in the default mkwrite function
1) Make file_update_time at the start before acquiring any lock
2) the condition page_offset(page) >= i_size_read(inode) should be
 changed to page_offset(page) > i_size_read
3) Move wait_on_page_writeback.

Signed-off-by: Namjae Jeon <namjae.jeon@samsung.com>
Signed-off-by: Amit Sahrawat <a.sahrawat@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
11 years agof2fs: use list_for_each_entry rather than list_for_each_entry_safe
majianpeng [Tue, 14 May 2013 12:06:46 +0000 (20:06 +0800)]
f2fs: use list_for_each_entry rather than list_for_each_entry_safe

We can do this, since now we use a global mutex, f2fs_stat_mutex to protect its
list operations.

Signed-off-by: Jianpeng Ma <majianpeng@gmail.com>
[Jaegeuk Kim: add description]
Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
11 years agof2fs: remove unecessary variable and code
Haicheng Li [Tue, 14 May 2013 10:20:28 +0000 (18:20 +0800)]
f2fs: remove unecessary variable and code

Code cleanup without behavior changed.

Signed-off-by: Haicheng Li <haicheng.li@linux.intel.com>
Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
11 years agof2fs, lockdep: annotate mutex_lock_all()
Peter Zijlstra [Thu, 16 May 2013 18:03:12 +0000 (20:03 +0200)]
f2fs, lockdep: annotate mutex_lock_all()

Majianpeng reported a lockdep splat for f2fs. It turns out mutex_lock_all()
acquires an array of locks (in global/local lock style).

Any such operation is always serialized using cp_mutex, therefore there is no
fs_lock[] lock-order issue; tell lockdep about this using the
mutex_lock_nest_lock() primitive.

Reported-by: majianpeng <majianpeng@gmail.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
11 years agof2fs: add debug msgs in the recovery routine
Jaegeuk Kim [Thu, 16 May 2013 06:04:49 +0000 (15:04 +0900)]
f2fs: add debug msgs in the recovery routine

This patch adds some trivial debugging messages in the recovery process.

Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
11 years agof2fs: update inode page after creation
Jaegeuk Kim [Mon, 20 May 2013 01:10:29 +0000 (10:10 +0900)]
f2fs: update inode page after creation

I found a bug when testing power-off-recovery as follows.

[Bug Scenario]
1. create a file
2. fsync the file
3. reboot w/o any sync
4. try to recover the file
 - found its fsync mark
 - found its dentry mark
   : try to recover its dentry
    - get its file name
    - get its parent inode number
     : here we got zero value

The reason why we get the wrong parent inode number is that we didn't
synchronize the inode page with its newly created inode information perfectly.

Especially, previous f2fs stores fi->i_pino and writes it to the cached
node page in a wrong order, which incurs the zero-valued i_pino during the
recovery.

So, this patch modifies the creation flow to fix the synchronization order of
inode page with its inode.

Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
11 years agof2fs: change get_new_data_page to pass a locked node page
Jaegeuk Kim [Mon, 20 May 2013 00:55:50 +0000 (09:55 +0900)]
f2fs: change get_new_data_page to pass a locked node page

This patch is for passing a locked node page to get_dnode_of_data.

Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
11 years agof2fs: skip get_node_page if locked node page is passed
Jaegeuk Kim [Mon, 20 May 2013 00:42:28 +0000 (09:42 +0900)]
f2fs: skip get_node_page if locked node page is passed

If get_dnode_of_data gets a locked node page, let's skip redundant
get_node_page calls.
This is for the futher enhancement.

Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
11 years agof2fs: remove unnecessary por_doing check
Jaegeuk Kim [Wed, 15 May 2013 23:57:43 +0000 (08:57 +0900)]
f2fs: remove unnecessary por_doing check

This por_doing check is totally not related to the recovery process.

Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
11 years agof2fs: fix BUG_ON during f2fs_evict_inode(dir)
Jaegeuk Kim [Wed, 15 May 2013 07:40:02 +0000 (16:40 +0900)]
f2fs: fix BUG_ON during f2fs_evict_inode(dir)

During the dentry recovery routine, recover_inode() triggers __f2fs_add_link
with its directory inode.

In the following scenario, a bug is captured.
 1. dir = f2fs_iget(pino)
 2. __f2fs_add_link(dir, name)
 3. iput(dir)
  -> f2fs_evict_inode() faces with BUG_ON(atomic_read(fi->dirty_dents))

Kernel BUG at ffffffffa01c0676 [verbose debug info unavailable]
[<ffffffffa01c0676>] f2fs_evict_inode+0x276/0x300 [f2fs]
Call Trace:
 [<ffffffff8118ea00>] evict+0xb0/0x1b0
 [<ffffffff8118f1c5>] iput+0x105/0x190
 [<ffffffffa01d2dac>] recover_fsync_data+0x3bc/0x1070 [f2fs]
 [<ffffffff81692e8a>] ? io_schedule+0xaa/0xd0
 [<ffffffff81690acb>] ? __wait_on_bit_lock+0x7b/0xc0
 [<ffffffff8111a0e7>] ? __lock_page+0x67/0x70
 [<ffffffff81165e21>] ? kmem_cache_alloc+0x31/0x140
 [<ffffffff8118a502>] ? __d_instantiate+0x92/0xf0
 [<ffffffff812a949b>] ? security_d_instantiate+0x1b/0x30
 [<ffffffff8118a5b4>] ? d_instantiate+0x54/0x70

This means that we should flush all the dentry pages between iget and iput().
But, during the recovery routine, it is unallowed due to consistency, so we
have to wait the whole recovery process.
And then, write_checkpoint flushes all the dirty dentry blocks, and nicely we
can put the stale dir inodes from the dirty_dir_inode_list.

Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
11 years agof2fs: fix por_doing variable coverage
Jaegeuk Kim [Wed, 15 May 2013 07:12:18 +0000 (16:12 +0900)]
f2fs: fix por_doing variable coverage

The reason of using sbi->por_doing is to alleviate data writes during the
recovery.
The find_fsync_dnodes() produces some dirty dentry pages, so we should
cover it too with sbi->por_doing.

Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
11 years agof2fs: remove redundant assignment
Jaegeuk Kim [Wed, 15 May 2013 01:49:13 +0000 (10:49 +0900)]
f2fs: remove redundant assignment

We don't need to assign a value redundantly.

Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
11 years agof2fs: fix the inconsistent state of data pages
Jaegeuk Kim [Sun, 12 May 2013 23:38:35 +0000 (08:38 +0900)]
f2fs: fix the inconsistent state of data pages

In get_lock_data_page, if there is a data race between get_dnode_of_data for
node and grab_cache_page for data, f2fs is able to face with the following
BUG_ON(dn.data_blkaddr == NEW_ADDR).

kernel BUG at /home/zeus/f2fs_test/src/fs/f2fs/data.c:251!
 [<ffffffffa044966c>] get_lock_data_page+0x1ec/0x210 [f2fs]
Call Trace:
 [<ffffffffa043b089>] f2fs_readdir+0x89/0x210 [f2fs]
 [<ffffffff811a0920>] ? fillonedir+0x100/0x100
 [<ffffffff811a0920>] ? fillonedir+0x100/0x100
 [<ffffffff811a07f8>] vfs_readdir+0xb8/0xe0
 [<ffffffff811a0b4f>] sys_getdents+0x8f/0x110
 [<ffffffff816d7999>] system_call_fastpath+0x16/0x1b

This bug is able to be occurred when the block address of the data block is
changed after f2fs_put_dnode().
In order to avoid that, this patch fixes the lock order of node and data
blocks in which the node block lock is covered by the data block lock.

Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
11 years agof2fs: fix inconsistency of block count during recovery
Jaegeuk Kim [Tue, 14 May 2013 06:47:43 +0000 (15:47 +0900)]
f2fs: fix inconsistency of block count during recovery

Currently f2fs recovers the dentry of fsynced files.
When power-off-recovery is conducted, this newly recovered inode should increase
node block count as well as inode block count.

This patch resolves this inconsistency that results in:

1. create a file
2. write data
3. fsync
4. reboot without sync
5. mount and recover the file
6. node block count is 1 and inode block count is 2
 : fall into the inconsistent state
7. unlink the file
 : trigger the following BUG_ON

------------[ cut here ]------------
kernel BUG at /home/zeus/f2fs_test/src/fs/f2fs/f2fs.h:716!
Call Trace:
 [<ffffffffa0344100>] ? get_node_page+0x50/0x1a0 [f2fs]
 [<ffffffffa0344bfc>] remove_inode_page+0x8c/0x100 [f2fs]
 [<ffffffffa03380f0>] ? f2fs_evict_inode+0x180/0x2d0 [f2fs]
 [<ffffffffa033812e>] f2fs_evict_inode+0x1be/0x2d0 [f2fs]
 [<ffffffff811c7a67>] evict+0xa7/0x1a0
 [<ffffffff811c82b5>] iput+0x105/0x190
 [<ffffffff811c2b30>] d_kill+0xe0/0x120
 [<ffffffff811c2c57>] dput+0xe7/0x1e0
 [<ffffffff811acc3d>] __fput+0x19d/0x2d0
 [<ffffffff811acd7e>] ____fput+0xe/0x10
 [<ffffffff81070645>] task_work_run+0xb5/0xe0
 [<ffffffff81002941>] do_notify_resume+0x71/0xb0
 [<ffffffff8175f14a>] int_signal+0x12/0x17

Reported-and-Tested-by: Chris Fries <C.Fries@motorola.com>
Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
11 years agoLinux 3.10-rc3
Linus Torvalds [Sun, 26 May 2013 23:00:47 +0000 (16:00 -0700)]
Linux 3.10-rc3

11 years agoipc/sem.c: Fix missing wakeups in do_smart_update_queue()
Manfred Spraul [Sun, 26 May 2013 09:08:52 +0000 (11:08 +0200)]
ipc/sem.c: Fix missing wakeups in do_smart_update_queue()

do_smart_update_queue() is called when an operation (semop,
semctl(SETVAL), semctl(SETALL), ...) modified the array.  It must check
which of the sleeping tasks can proceed.

do_smart_update_queue() missed a few wakeups:
 - if a sleeping complex op was completed, then all per-semaphore queues
   must be scanned - not only those that were modified by *sops
 - if a sleeping simple op proceeded, then the global queue must be
   scanned again

And:
 - the test for "|sops == NULL) before scanning the global queue is not
   required: If the global queue is empty, then it doesn't need to be
   scanned - regardless of the reason for calling do_smart_update_queue()

The patch is not optimized, i.e.  even completing a wait-for-zero
operation causes a rescan.  This is done to keep the patch as simple as
possible.

Signed-off-by: Manfred Spraul <manfred@colorfullife.com>
Acked-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agoMerge tag 'nfs-for-3.10-3' of git://git.linux-nfs.org/projects/trondmy/linux-nfs
Linus Torvalds [Sun, 26 May 2013 19:33:05 +0000 (12:33 -0700)]
Merge tag 'nfs-for-3.10-3' of git://git.linux-nfs.org/projects/trondmy/linux-nfs

Pull NFS client bugfixes from Trond Myklebust:

 - Stable fix to prevent an rpc_task wakeup race
 - Fix a NFSv4.1 session drain deadlock
 - Fix a NFSv4/v4.1 mount regression when not running rpc.gssd
 - Ensure auth_gss pipe detection works in namespaces
 - Fix SETCLIENTID fallback if rpcsec_gss is not available

* tag 'nfs-for-3.10-3' of git://git.linux-nfs.org/projects/trondmy/linux-nfs:
  NFS: Fix SETCLIENTID fallback if GSS is not available
  SUNRPC: Prevent an rpc_task wakeup race
  NFSv4.1 Fix a pNFS session draining deadlock
  SUNRPC: Convert auth_gss pipe detection to work in namespaces
  SUNRPC: Faster detection if gssd is actually running
  SUNRPC: Fix a bug in gss_create_upcall

11 years agoMerge tag 'edac_fixes_for_3.10' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sun, 26 May 2013 16:52:26 +0000 (09:52 -0700)]
Merge tag 'edac_fixes_for_3.10' of git://git./linux/kernel/git/bp/bp

Pull amd64 edac fix from Borislav Petkov:
 "A sysfs file permissions correction"

* tag 'edac_fixes_for_3.10' of git://git.kernel.org/pub/scm/linux/kernel/git/bp/bp:
  amd64_edac: Fix bogus sysfs file permissions

11 years agoMerge branch 'parisc-for-3.10' of git://git.kernel.org/pub/scm/linux/kernel/git/delle...
Linus Torvalds [Sun, 26 May 2013 16:36:31 +0000 (09:36 -0700)]
Merge branch 'parisc-for-3.10' of git://git./linux/kernel/git/deller/parisc-linux

Pull parisc fixes from Helge Deller:
 "This time we made the kernel- and interruption stack allocation
  reentrant which fixed some strange kernel crashes (specifically
  protection ID traps).

  Furthemore this patchset fixes the interrupt stack in UP and SMP
  configurations by using native locking instructions.  And finally
  usage of floating point calculations on parisc were disabled in the
  MPILIB."

* 'parisc-for-3.10' of git://git.kernel.org/pub/scm/linux/kernel/git/deller/parisc-linux:
  parisc: fix irq stack on UP and SMP
  parisc/superio: Use module_pci_driver to register driver
  parisc: make interrupt and interruption stack allocation reentrant
  parisc: show number of FPE and unaligned access handler calls in /proc/interrupts
  parisc: add additional parisc git tree to MAINTAINERS file
  parisc: use PAGE_SHIFT instead of hardcoded value 12 in pacache.S
  parisc: add rp5470 entry to machine database
  MPILIB: disable usage of floating point registers on parisc

11 years agoMerge tag 'for-linus-v3.10-rc3' of git://oss.sgi.com/xfs/xfs
Linus Torvalds [Sun, 26 May 2013 16:35:02 +0000 (09:35 -0700)]
Merge tag 'for-linus-v3.10-rc3' of git://oss.sgi.com/xfs/xfs

Pull xfs fixes from Ben Myers:
 "Here are fixes for corruption on 512 byte filesystems, a rounding
  error, a use-after-free, some flags to fix lockdep reports, and
  several fixes related to CRCs.  We have a somewhat larger post -rc1
  queue than usual due to fixes related to the CRC feature we merged for
  3.10:

   - Fix for corruption with FSX on 512 byte blocksize filesystems
   - Fix rounding error in xfs_free_file_space
   - Fix use-after-free with extent free intents
   - Add several missing KM_NOFS flags to fix lockdep reports
   - Several fixes for CRC related code"

* tag 'for-linus-v3.10-rc3' of git://oss.sgi.com/xfs/xfs:
  xfs: remote attribute lookups require the value length
  xfs: xfs_attr_shortform_allfit() does not handle attr3 format.
  xfs: xfs_da3_node_read_verify() doesn't handle XFS_ATTR3_LEAF_MAGIC
  xfs: fix missing KM_NOFS tags to keep lockdep happy
  xfs: Don't reference the EFI after it is freed
  xfs: fix rounding in xfs_free_file_space
  xfs: fix sub-page blocksize data integrity writes

11 years agoMerge tag 'for-v3.10-fixes' of git://git.infradead.org/battery-2.6
Linus Torvalds [Sun, 26 May 2013 03:32:49 +0000 (20:32 -0700)]
Merge tag 'for-v3.10-fixes' of git://git.infradead.org/battery-2.6

Pull bettery fixes from Anton Vorontsov:
 "Last minute one-liners: wrong kfree usage fix, module alias fixup and
  kconfig adjustments"

* tag 'for-v3.10-fixes' of git://git.infradead.org/battery-2.6:
  pm2301_charger: Fix module alias prefix
  wm831x_backup: Fix wrong kfree call for devdata->backup.name
  bq27x00: Fix I2C dependency in KConfig
  lp8788-charger: Fix kconfig dependency

11 years agoMerge tag 'pm+acpi-3.10-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael...
Linus Torvalds [Sun, 26 May 2013 03:32:00 +0000 (20:32 -0700)]
Merge tag 'pm+acpi-3.10-rc3' of git://git./linux/kernel/git/rafael/linux-pm

Pull power management and ACPI fixes from Rafael Wysocki:

 - Additional CPU ID for the intel_pstate driver from Dirk Brandewie.

 - More cpufreq fixes related to ARM big.LITTLE support and locking from
   Viresh Kumar.

 - VIA C7 cpufreq build fix from RafaÅ‚ Bilski.

 - ACPI power management fix making it possible to use device power
   states regardless of the CONFIG_PM setting from Rafael J Wysocki.

 - New ACPI video blacklist item from Bastian Triller.

* tag 'pm+acpi-3.10-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
  ACPI / video: Add "Asus UL30A" to ACPI video detect blacklist
  cpufreq: arm_big_little_dt: Instantiate as platform_driver
  cpufreq: arm_big_little_dt: Register driver only if DT has valid data
  cpufreq / e_powersaver: Fix linker error when ACPI processor is a module
  cpufreq / intel_pstate: Add additional supported CPU ID
  cpufreq: Drop rwsem lock around CPUFREQ_GOV_POLICY_EXIT
  ACPI / PM: Allow device power states to be used for CONFIG_PM unset

11 years agoMerge branch 'fixes' of git://git.infradead.org/users/vkoul/slave-dma
Linus Torvalds [Sun, 26 May 2013 03:30:31 +0000 (20:30 -0700)]
Merge branch 'fixes' of git://git.infradead.org/users/vkoul/slave-dma

Pull slave-dma fixes from Vinod Koul:
 "We have two patches from Andy & Rafael fixing the Lynxpoint dma"

* 'fixes' of git://git.infradead.org/users/vkoul/slave-dma:
  ACPI / LPSS: register clock device for Lynxpoint DMA properly
  dma: acpi-dma: parse CSRT to extract additional resources

11 years agoscore: remove redundant kcore_list entries
Kyle McMartin [Sat, 25 May 2013 16:54:08 +0000 (12:54 -0400)]
score: remove redundant kcore_list entries

kcore_vmalloc is in fs/proc/kcore.c and kcore_mem is unused across
the tree. Noticed while grepping the tree for some other kcore stuff.

(score looks pretty unmaintained to me.)

Signed-off-by: Kyle McMartin <kyle@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agoMerge tag 'arc-v3.10-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/vgupta/arc
Linus Torvalds [Sat, 25 May 2013 17:06:20 +0000 (10:06 -0700)]
Merge tag 'arc-v3.10-fixes' of git://git./linux/kernel/git/vgupta/arc

Pull ARC fixes from Vineet Gupta:

 - Fallouts/wreckage of Cache Flush optimizations / aliasing dcache
   support

 - Fix for an interesting bug where piped input to grep was getting
   mysteriously clobbered

* tag 'arc-v3.10-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/vgupta/arc:
  ARC: lazy dcache flush broke gdb in non-aliasing configs
  ARC: Use enough bits for determining page's cache color
  ARC: Brown paper bag bug in macro for checking cache color
  ARC: copy_(to|from)_user() to honor usermode-access permissions
  ARC: [mm] Prevent stray dcache lines after__sync_icache_dcach()
  ARC: [TB10x] Remove redundant abilis,simple-pinctrl mechanism

11 years agoMerge branch 'fixes' of git://git.linaro.org/people/rmk/linux-arm
Linus Torvalds [Sat, 25 May 2013 17:05:24 +0000 (10:05 -0700)]
Merge branch 'fixes' of git://git.linaro.org/people/rmk/linux-arm

Pull ARM fixes from Russell King:
 "Just three this time, all really quite small"

* 'fixes' of git://git.linaro.org/people/rmk/linux-arm:
  ARM: 7729/1: vfp: ensure VFP_arch is non-zero when VFP is not supported
  ARM: 7727/1: remove the .vm_mm value from gate_vma
  ARM: 7723/1: crypto: sha1-armv4-large.S: fix SP handling

11 years agoARC: lazy dcache flush broke gdb in non-aliasing configs
Vineet Gupta [Sat, 25 May 2013 08:34:25 +0000 (14:04 +0530)]
ARC: lazy dcache flush broke gdb in non-aliasing configs

gdbserver inserting a breakpoint ends up calling copy_user_page() for a
code page. The generic version of which (non-aliasing config) didn't set
the PG_arch_1 bit hence update_mmu_cache() didn't sync dcache/icache for
corresponding dynamic loader code page - causing garbade to be executed.

So now aliasing versions of copy_user_highpage()/clear_page() are made
default. There is no significant overhead since all of special alias
handling code is compiled out for non-aliasing build

Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
11 years agoMerge branch 'akpm' (incoming from Andrew Morton)
Linus Torvalds [Sat, 25 May 2013 01:12:15 +0000 (18:12 -0700)]
Merge branch 'akpm' (incoming from Andrew Morton)

Merge fixes from Andrew Morton:
 "A bunch of fixes and one simple fbdev driver which missed the merge
  window because people will still talking about it (to no great
  effect)."

* emailed patches from Andrew Morton <akpm@linux-foundation.org>: (30 commits)
  aio: fix kioctx not being freed after cancellation at exit time
  mm/pagewalk.c: walk_page_range should avoid VM_PFNMAP areas
  drivers/rtc/rtc-max8998.c: check for pdata presence before dereferencing
  ocfs2: goto out_unlock if ocfs2_get_clusters_nocache() failed in ocfs2_fiemap()
  random: fix accounting race condition with lockless irq entropy_count update
  drivers/char/random.c: fix priming of last_data
  mm/memory_hotplug.c: fix printk format warnings
  nilfs2: fix issue of nilfs_set_page_dirty() for page at EOF boundary
  drivers/block/brd.c: fix brd_lookup_page() race
  fbdev: FB_GOLDFISH should depend on HAS_DMA
  drivers/rtc/rtc-pl031.c: pass correct pointer to free_irq()
  auditfilter.c: fix kernel-doc warnings
  aio: fix io_getevents documentation
  revert "selftest: add simple test for soft-dirty bit"
  drivers/leds/leds-ot200.c: fix error caused by shifted mask
  mm/THP: use pmd_populate() to update the pmd with pgtable_t pointer
  linux/kernel.h: fix kernel-doc warning
  mm compaction: fix of improper cache flush in migration code
  rapidio/tsi721: fix bug in MSI interrupt handling
  hfs: avoid crash in hfs_bnode_create
  ...

11 years agoMerge tag 'fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/arm...
Linus Torvalds [Fri, 24 May 2013 23:27:37 +0000 (16:27 -0700)]
Merge tag 'fixes-for-linus' of git://git./linux/kernel/git/arm/arm-soc

Pull ARM SoC fixes from Olof Johansson:
 "We didn't have any fixes sent up for -rc2, so this is a slightly
  larger batch.  A bit all over the place platform-wise; OMAP, at91,
  marvell, renesas, sunxi, ux500, etc.

  I tried to summarize highlights but there isn't a whole lot to point
  out.  Lots of little things fixed all over.  A couple of defconfig
  updates due to new/changing options."

* tag 'fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc: (44 commits)
  ARM: at91/sama5: fix incorrect PMC pcr div definition
  ARM: at91/dt: fix macb pinctrl_macb_rmii_mii_alt definition
  ARM: at91: at91sam9n12: move external irq declatation to DT
  ARM: shmobile: marzen: Use error values in usb_power_*
  ARM: tegra: defconfig fixes
  ARM: nomadik: fix IRQ assignment for SMC ethernet
  ARM: vt8500: Add missing NULL terminator in dt_compat
  clk: tegra: add ac97 controller clock
  clk: tegra: remove USB from clk init table
  ARM: dts: mvebu: Fix wrong the address reg value for the L2-cache node
  ARM: plat-orion: Fix num_resources and id for ge10 and ge11
  ARM: OMAP2+: hwmod: Remove sysc slave idle and auto idle apis
  SERIAL: OMAP: Remove the slave idle handling from the driver
  ARM: OMAP2+: serial: Remove the un-used slave idle hooks
  ARM: OMAP2+: hwmod-data: UART IP needs software control to manage sidle modes
  ARM: OMAP2+: hwmod: Add a new flag to handle SIDLE in SWSUP only in active
  ARM: OMAP2+: hwmod: Fix sidle programming in _enable_sysc()/_idle_sysc()
  arm: mvebu: fix the 'ranges' property to handle PCIe
  ARM: mvebu: select ARCH_REQUIRE_GPIOLIB for mvebu platform
  ARM: AM33XX: Add missing .clkdm_name to clkdiv32k_ick clock
  ...

11 years agoaio: fix kioctx not being freed after cancellation at exit time
Benjamin LaHaise [Fri, 24 May 2013 22:55:38 +0000 (15:55 -0700)]
aio: fix kioctx not being freed after cancellation at exit time

The recent changes overhauling fs/aio.c introduced a bug that results in
the kioctx not being freed when outstanding kiocbs are cancelled at
exit_aio() time.  Specifically, a kiocb that is cancelled has its
completion events discarded by batch_complete_aio(), which then fails to
wake up the process stuck in free_ioctx().  Fix this by modifying the
wait_event() condition in free_ioctx() appropriately.

This patch was tested with the cancel operation in the thread based code
posted yesterday.

[akpm@linux-foundation.org: fix build]
Signed-off-by: Benjamin LaHaise <bcrl@kvack.org>
Signed-off-by: Kent Overstreet <koverstreet@google.com>
Cc: Kent Overstreet <koverstreet@google.com>
Cc: Josh Boyer <jwboyer@redhat.com>
Cc: Zach Brown <zab@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agomm/pagewalk.c: walk_page_range should avoid VM_PFNMAP areas
Cliff Wickman [Fri, 24 May 2013 22:55:36 +0000 (15:55 -0700)]
mm/pagewalk.c: walk_page_range should avoid VM_PFNMAP areas

A panic can be caused by simply cat'ing /proc/<pid>/smaps while an
application has a VM_PFNMAP range.  It happened in-house when a
benchmarker was trying to decipher the memory layout of his program.

/proc/<pid>/smaps and similar walks through a user page table should not
be looking at VM_PFNMAP areas.

Certain tests in walk_page_range() (specifically split_huge_page_pmd())
assume that all the mapped PFN's are backed with page structures.  And
this is not usually true for VM_PFNMAP areas.  This can result in panics
on kernel page faults when attempting to address those page structures.

There are a half dozen callers of walk_page_range() that walk through a
task's entire page table (as N.  Horiguchi pointed out).  So rather than
change all of them, this patch changes just walk_page_range() to ignore
VM_PFNMAP areas.

The logic of hugetlb_vma() is moved back into walk_page_range(), as we
want to test any vma in the range.

VM_PFNMAP areas are used by:
- graphics memory manager   gpu/drm/drm_gem.c
- global reference unit     sgi-gru/grufile.c
- sgi special memory        char/mspec.c
- and probably several out-of-tree modules

[akpm@linux-foundation.org: remove now-unused hugetlb_vma() stub]
Signed-off-by: Cliff Wickman <cpw@sgi.com>
Reviewed-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: David Sterba <dsterba@suse.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: KOSAKI Motohiro <kosaki.motohiro@gmail.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agodrivers/rtc/rtc-max8998.c: check for pdata presence before dereferencing
Tomasz Figa [Fri, 24 May 2013 22:55:35 +0000 (15:55 -0700)]
drivers/rtc/rtc-max8998.c: check for pdata presence before dereferencing

Currently the driver can crash with a NULL pointer dereference if no
pdata is provided, despite of successful registration of the MFD part.
This patch fixes the problem by adding a NULL check before dereferencing
the pdata pointer.

Signed-off-by: Tomasz Figa <t.figa@samsung.com>
Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com>
Cc: Sachin Kamat <sachin.kamat@linaro.org>
Reviewed-by: Jingoo Han <jg1.han@samsung.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agoocfs2: goto out_unlock if ocfs2_get_clusters_nocache() failed in ocfs2_fiemap()
Joseph Qi [Fri, 24 May 2013 22:55:34 +0000 (15:55 -0700)]
ocfs2: goto out_unlock if ocfs2_get_clusters_nocache() failed in ocfs2_fiemap()

Last time we found there is lock/unlock bug in ocfs2_file_aio_write, and
then we did a thorough search for all lock resources in
ocfs2_inode_info, including rw, inode and open lockres and found this
bug.  My kernel version is 3.0.13, and it is also in the lastest version
3.9.  In ocfs2_fiemap, once ocfs2_get_clusters_nocache failed, it should
goto out_unlock instead of out, because we need release buffer head, up
read alloc sem and unlock inode.

Signed-off-by: Joseph Qi <joseph.qi@huawei.com>
Reviewed-by: Jie Liu <jeff.liu@oracle.com>
Cc: Mark Fasheh <mfasheh@suse.com>
Cc: Joel Becker <jlbec@evilplan.org>
Acked-by: Sunil Mushran <sunil.mushran@gmail.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agorandom: fix accounting race condition with lockless irq entropy_count update
Jiri Kosina [Fri, 24 May 2013 22:55:33 +0000 (15:55 -0700)]
random: fix accounting race condition with lockless irq entropy_count update

Commit 902c098a3663 ("random: use lockless techniques in the interrupt
path") turned IRQ path from being spinlock protected into lockless
cmpxchg-retry update.

That commit removed r->lock serialization between crediting entropy bits
from IRQ context and accounting when extracting entropy on userspace
read path, but didn't turn the r->entropy_count reads/updates in
account() to use cmpxchg as well.

It has been observed, that under certain circumstances this leads to
read() on /dev/urandom to return 0 (EOF), as r->entropy_count gets
corrupted and becomes negative, which in turn results in propagating 0
all the way from account() to the actual read() call.

Convert the accounting code to be the proper lockless counterpart of
what has been partially done by 902c098a3663.

Signed-off-by: Jiri Kosina <jkosina@suse.cz>
Cc: Theodore Ts'o <tytso@mit.edu>
Cc: Greg KH <greg@kroah.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agodrivers/char/random.c: fix priming of last_data
Jarod Wilson [Fri, 24 May 2013 22:55:31 +0000 (15:55 -0700)]
drivers/char/random.c: fix priming of last_data

Commit ec8f02da9ea5 ("random: prime last_data value per fips
requirements") added priming of last_data per fips requirements.

Unfortuantely, it did so in a way that can lead to multiple threads all
incrementing nbytes, but only one actually doing anything with the extra
data, which leads to some fun random corruption and panics.

The fix is to simply do everything needed to prime last_data in a single
shot, so there's no window for multiple cpus to increment nbytes -- in
fact, we won't even increment or decrement nbytes anymore, we'll just
extract the needed EXTRACT_SIZE one time per pool and then carry on with
the normal routine.

All these changes have been tested across multiple hosts and
architectures where panics were previously encoutered.  The code changes
are are strictly limited to areas only touched when when booted in fips
mode.

This change should also go into 3.8-stable, to make the myriads of fips
users on 3.8.x happy.

Signed-off-by: Jarod Wilson <jarod@redhat.com>
Tested-by: Jan Stancek <jstancek@redhat.com>
Tested-by: Jan Stodola <jstodola@redhat.com>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Matt Mackall <mpm@selenic.com>
Cc: "Theodore Ts'o" <tytso@mit.edu>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agomm/memory_hotplug.c: fix printk format warnings
Randy Dunlap [Fri, 24 May 2013 22:55:30 +0000 (15:55 -0700)]
mm/memory_hotplug.c: fix printk format warnings

Fix printk format warnings in mm/memory_hotplug.c by using "%pa":

  mm/memory_hotplug.c: warning: format '%llx' expects argument of type 'long long unsigned int', but argument 2 has type 'resource_size_t' [-Wformat]
  mm/memory_hotplug.c: warning: format '%llx' expects argument of type 'long long unsigned int', but argument 3 has type 'resource_size_t' [-Wformat]

Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Reported-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agonilfs2: fix issue of nilfs_set_page_dirty() for page at EOF boundary
Ryusuke Konishi [Fri, 24 May 2013 22:55:29 +0000 (15:55 -0700)]
nilfs2: fix issue of nilfs_set_page_dirty() for page at EOF boundary

nilfs2: fix issue of nilfs_set_page_dirty for page at EOF boundary

DESCRIPTION:
 There are use-cases when NILFS2 file system (formatted with block size
lesser than 4 KB) can be remounted in RO mode because of encountering of
"broken bmap" issue.

The issue was reported by Anthony Doggett <Anthony2486@interfaces.org.uk>:
 "The machine I've been trialling nilfs on is running Debian Testing,
  Linux version 3.2.0-4-686-pae (debian-kernel@lists.debian.org) (gcc
  version 4.6.3 (Debian 4.6.3-14) ) #1 SMP Debian 3.2.35-2), but I've
  also reproduced it (identically) with Debian Unstable amd64 and Debian
  Experimental (using the 3.8-trunk kernel).  The problematic partitions
  were formatted with "mkfs.nilfs2 -b 1024 -B 8192"."

SYMPTOMS:
(1) System log contains error messages likewise:

    [63102.496756] nilfs_direct_assign: invalid pointer: 0
    [63102.496786] NILFS error (device dm-17): nilfs_bmap_assign: broken bmap (inode number=28)
    [63102.496798]
    [63102.524403] Remounting filesystem read-only

(2) The NILFS2 file system is remounted in RO mode.

REPRODUSING PATH:
(1) Create volume group with name "unencrypted" by means of vgcreate utility.
(2) Run script (prepared by Anthony Doggett <Anthony2486@interfaces.org.uk>):

----------------[BEGIN SCRIPT]--------------------

VG=unencrypted
lvcreate --size 2G --name ntest $VG
mkfs.nilfs2 -b 1024 -B 8192 /dev/mapper/$VG-ntest
mkdir /var/tmp/n
mkdir /var/tmp/n/ntest
mount /dev/mapper/$VG-ntest /var/tmp/n/ntest
mkdir /var/tmp/n/ntest/thedir
cd /var/tmp/n/ntest/thedir
sleep 2
date
darcs init
sleep 2
dmesg|tail -n 5
date
darcs whatsnew || true
date
sleep 2
dmesg|tail -n 5
----------------[END SCRIPT]--------------------

REPRODUCIBILITY: 100%

INVESTIGATION:
As it was discovered, the issue takes place during segment
construction after executing such sequence of user-space operations:

  open("_darcs/index", O_RDWR|O_CREAT|O_NOCTTY, 0666) = 7
  fstat(7, {st_mode=S_IFREG|0644, st_size=0, ...}) = 0
  ftruncate(7, 60)

The error message "NILFS error (device dm-17): nilfs_bmap_assign: broken
bmap (inode number=28)" takes place because of trying to get block
number for third block of the file with logical offset #3072 bytes.  As
it is possible to see from above output, the file has 60 bytes of the
whole size.  So, it is enough one block (1 KB in size) allocation for
the whole file.  Trying to operate with several blocks instead of one
takes place because of discovering several dirty buffers for this file
in nilfs_segctor_scan_file() method.

The root cause of this issue is in nilfs_set_page_dirty function which
is called just before writing to an mmapped page.

When nilfs_page_mkwrite function handles a page at EOF boundary, it
fills hole blocks only inside EOF through __block_page_mkwrite().

The __block_page_mkwrite() function calls set_page_dirty() after filling
hole blocks, thus nilfs_set_page_dirty function (=
a_ops->set_page_dirty) is called.  However, the current implementation
of nilfs_set_page_dirty() wrongly marks all buffers dirty even for page
at EOF boundary.

As a result, buffers outside EOF are inconsistently marked dirty and
queued for write even though they are not mapped with nilfs_get_block
function.

FIX:
This modifies nilfs_set_page_dirty() not to mark hole blocks dirty.

Thanks to Vyacheslav Dubeyko for his effort on analysis and proposals
for this issue.

Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>
Reported-by: Anthony Doggett <Anthony2486@interfaces.org.uk>
Reported-by: Vyacheslav Dubeyko <slava@dubeyko.com>
Cc: Vyacheslav Dubeyko <slava@dubeyko.com>
Tested-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agodrivers/block/brd.c: fix brd_lookup_page() race
Brian Behlendorf [Fri, 24 May 2013 22:55:28 +0000 (15:55 -0700)]
drivers/block/brd.c: fix brd_lookup_page() race

The index on the page must be set before it is inserted in the radix
tree.  Otherwise there is a small race which can occur during lookup
where the page can be found with the incorrect index.  This will trigger
the BUG_ON() in brd_lookup_page().

Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reported-by: Chris Wedgwood <cw@f00f.org>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agofbdev: FB_GOLDFISH should depend on HAS_DMA
Geert Uytterhoeven [Fri, 24 May 2013 22:55:27 +0000 (15:55 -0700)]
fbdev: FB_GOLDFISH should depend on HAS_DMA

If NO_DMA=y:

  drivers/built-in.o: In function `goldfish_fb_remove':
  drivers/video/goldfishfb.c:301: undefined reference to `dma_free_coherent'
  drivers/built-in.o: In function `goldfish_fb_probe':
  drivers/video/goldfishfb.c:247: undefined reference to `dma_alloc_coherent'
  drivers/video/goldfishfb.c:280: undefined reference to `dma_free_coherent'

Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Florian Tobias Schandinat <FlorianSchandinat@gmx.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agodrivers/rtc/rtc-pl031.c: pass correct pointer to free_irq()
Lars-Peter Clausen [Fri, 24 May 2013 22:55:26 +0000 (15:55 -0700)]
drivers/rtc/rtc-pl031.c: pass correct pointer to free_irq()

free_irq() expects the same pointer that was passed to request_irq(),
otherwise the IRQ is not freed.

The issue was found using the following coccinelle script:

  <smpl>
  @r1@
  type T;
  T devid;
  @@
  request_irq(..., devid)

  @r2@
  type r1.T;
  T devid;
  position p;
  @@
  free_irq@p(..., devid)

  @@
  position p != r2.p;
  @@
  *free_irq@p(...)
  </smpl>

Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
Cc: Srinidhi Kasagar <srinidhi.kasagar@stericsson.com>
Cc: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agoauditfilter.c: fix kernel-doc warnings
Randy Dunlap [Fri, 24 May 2013 22:55:25 +0000 (15:55 -0700)]
auditfilter.c: fix kernel-doc warnings

Fix kernel-doc warnings in kernel/auditfilter.c:

  Warning(kernel/auditfilter.c:1029): Excess function parameter 'loginuid' description in 'audit_receive_filter'
  Warning(kernel/auditfilter.c:1029): Excess function parameter 'sessionid' description in 'audit_receive_filter'
  Warning(kernel/auditfilter.c:1029): Excess function parameter 'sid' description in 'audit_receive_filter'

Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Cc: Eric Paris <eparis@redhat.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agoaio: fix io_getevents documentation
Jeff Moyer [Fri, 24 May 2013 22:55:24 +0000 (15:55 -0700)]
aio: fix io_getevents documentation

In reviewing man pages, I noticed that io_getevents is documented to
update the timeout that gets passed into the library call.  This doesn't
happen in kernel space or in the library (even though it's documented to
do so in both places).  Unless there is objection, I'd like to fix the
comments/docs to match the code (I will also update the man page upon
consensus).

Signed-off-by: Jeff Moyer <jmoyer@redhat.com>
Signed-off-by: Benjamin LaHaise <bcrl@kvack.org>
Acked-by: Cyril Hrubis <chrubis@suse.cz>
Acked-by: Michael Kerrisk <mtk.manpages@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agorevert "selftest: add simple test for soft-dirty bit"
Andrew Morton [Fri, 24 May 2013 22:55:23 +0000 (15:55 -0700)]
revert "selftest: add simple test for soft-dirty bit"

Revert commit 58c7be84fec8 ("selftest: add simple test for soft-dirty
bit").  This is the self test for Pavel's pagemap2 patches which didn't
actually get merged.

Reported-by: Pavel Emelyanov <xemul@parallels.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agodrivers/leds/leds-ot200.c: fix error caused by shifted mask
Christian Gmeiner [Fri, 24 May 2013 22:55:22 +0000 (15:55 -0700)]
drivers/leds/leds-ot200.c: fix error caused by shifted mask

During the development of this driver an in-house register documentation
was used.  The last week some integration tests were done and this
problem was found.  It turned out that the released register
documentation is wrong.

The fix is very simple: shift all masks by one.

Signed-off-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Cc: Bryan Wu <cooloney@gmail.com>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agomm/THP: use pmd_populate() to update the pmd with pgtable_t pointer
Aneesh Kumar K.V [Fri, 24 May 2013 22:55:21 +0000 (15:55 -0700)]
mm/THP: use pmd_populate() to update the pmd with pgtable_t pointer

We should not use set_pmd_at to update pmd_t with pgtable_t pointer.
set_pmd_at is used to set pmd with huge pte entries and architectures
like ppc64, clear few flags from the pte when saving a new entry.
Without this change we observe bad pte errors like below on ppc64 with
THP enabled.

  BUG: Bad page map in process ld mm=0xc000001ee39f4780 pte:7fc3f37848000001 pmd:c000001ec0000000

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Reviewed-by: Andrea Arcangeli <aarcange@redhat.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agolinux/kernel.h: fix kernel-doc warning
Randy Dunlap [Fri, 24 May 2013 22:55:20 +0000 (15:55 -0700)]
linux/kernel.h: fix kernel-doc warning

Fix kernel-doc warning in <linux/kernel.h>:

  Warning(include/linux/kernel.h:590): No description found for parameter 'ip'

scripts/kernel-doc cannot handle macros, functions, or function
prototypes between the function or macro that is being documented and
its definition, so move these prototypes above the function that is
being documented.

Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Cc: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agomm compaction: fix of improper cache flush in migration code
Leonid Yegoshin [Fri, 24 May 2013 22:55:18 +0000 (15:55 -0700)]
mm compaction: fix of improper cache flush in migration code

Page 'new' during MIGRATION can't be flushed with flush_cache_page().
Using flush_cache_page(vma, addr, pfn) is justified only if the page is
already placed in process page table, and that is done right after
flush_cache_page().  But without it the arch function has no knowledge
of process PTE and does nothing.

Besides that, flush_cache_page() flushes an application cache page, but
the kernel has a different page virtual address and dirtied it.

Replace it with flush_dcache_page(new) which is the proper usage.

The old page is flushed in try_to_unmap_one() before migration.

This bug takes place in Sead3 board with M14Kc MIPS CPU without cache
aliasing (but Harvard arch - separate I and D cache) in tight memory
environment (128MB) each 1-3days on SOAK test.  It fails in cc1 during
kernel build (SIGILL, SIGBUS, SIGSEG) if CONFIG_COMPACTION is switched
ON.

Signed-off-by: Leonid Yegoshin <Leonid.Yegoshin@imgtec.com>
Cc: Leonid Yegoshin <yegoshin@mips.com>
Acked-by: Rik van Riel <riel@redhat.com>
Cc: Michal Hocko <mhocko@suse.cz>
Acked-by: Mel Gorman <mgorman@suse.de>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Russell King <rmk@arm.linux.org.uk>
Cc: David Miller <davem@davemloft.net>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agorapidio/tsi721: fix bug in MSI interrupt handling
Alexandre Bounine [Fri, 24 May 2013 22:55:17 +0000 (15:55 -0700)]
rapidio/tsi721: fix bug in MSI interrupt handling

Fix bug in MSI interrupt handling which causes loss of event
notifications.

Typical indication of lost MSI interrupts are stalled message and
doorbell transfers between RapidIO endpoints.  To avoid loss of MSI
interrupts all interrupts from the device must be disabled on entering
the interrupt handler routine and re-enabled when exiting it.
Re-enabling device interrupts will trigger new MSI message(s) if Tsi721
registered new events since entering interrupt handler routine.

This patch is applicable to kernel versions starting from v3.2.

Signed-off-by: Alexandre Bounine <alexandre.bounine@idt.com>
Cc: Matt Porter <mporter@kernel.crashing.org>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agohfs: avoid crash in hfs_bnode_create
Jeff Mahoney [Fri, 24 May 2013 22:55:16 +0000 (15:55 -0700)]
hfs: avoid crash in hfs_bnode_create

Commit 634725a92938 ("hfs: cleanup HFS+ prints") removed the BUG_ON in
hfs_bnode_create in hfsplus.  This patch removes it from the hfs version
and avoids an fsfuzzer crash.

Signed-off-by: Jeff Mahoney <jeffm@suse.com>
Acked-by: Jeff Mahoney <jeffm@suse.com>
Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Cc: Vyacheslav Dubeyko <slava@dubeyko.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agomm: memcg: remove incorrect VM_BUG_ON for swap cache pages in uncharge
Johannes Weiner [Fri, 24 May 2013 22:55:15 +0000 (15:55 -0700)]
mm: memcg: remove incorrect VM_BUG_ON for swap cache pages in uncharge

Commit 0c59b89c81ea ("mm: memcg: push down PageSwapCache check into
uncharge entry functions") added a VM_BUG_ON() on PageSwapCache in the
uncharge path after checking that page flag once, assuming that the
state is stable in all paths, but this is not the case and the condition
triggers in user environments.  An uncharge after the last page table
reference to the page goes away can race with reclaim adding the page to
swap cache.

Swap cache pages are usually uncharged when they are freed after
swapout, from a path that also handles swap usage accounting and memcg
lifetime management.  However, since the last page table reference is
gone and thus no references to the swap slot left, the swap slot will be
freed shortly when reclaim attempts to write the page to disk.  The
whole swap accounting is not even necessary.

So while the race condition for which this VM_BUG_ON was added is real
and actually existed all along, there are no negative effects.  Remove
the VM_BUG_ON again.

Reported-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Reported-by: Lingzhu Xiang <lxiang@redhat.com>
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Hugh Dickins <hughd@google.com>
Acked-by: Michal Hocko <mhocko@suse.cz>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agodrivers/video: implement a simple framebuffer driver
Stephen Warren [Fri, 24 May 2013 22:55:13 +0000 (15:55 -0700)]
drivers/video: implement a simple framebuffer driver

A simple frame-buffer describes a raw memory region that may be rendered
to, with the assumption that the display hardware has already been set
up to scan out from that buffer.

This is useful in cases where a bootloader exists and has set up the
display hardware, but a Linux driver doesn't yet exist for the display
hardware.

Examples use-cases include:

* The built-in LCD panels on the Samsung ARM chromebook, and Tegra
  devices, and likely many other ARM or embedded systems.  These cannot
  yet be supported using a full graphics driver, since the panel control
  should be provided by the CDF (Common Display Framework), which has been
  stuck in design/review for quite some time.  One could support these
  panels using custom SoC-specific code, but there is a desire to use
  common infra-structure rather than having each SoC vendor invent their
  own code, hence the desire to wait for CDF.

* Hardware for which a full graphics driver is not yet available, and
  the path to obtain one upstream isn't yet clear.  For example, the
  Raspberry Pi.

* Any hardware in early stages of upstreaming, before a full graphics
  driver has been tackled.  This driver can provide a graphical boot
  console (even full X support) much earlier in the upstreaming process,
  thus making new SoC or board support more generally useful earlier.

[akpm@linux-foundation.org: make simplefb_formats[] static]
Signed-off-by: Stephen Warren <swarren@wwwdotorg.org>
Cc: Arnd Bergmann <arnd@arndb.de>
Acked-by: Olof Johansson <olof@lixom.net>
Cc: Rob Clark <robclark@gmail.com>
Cc: Florian Tobias Schandinat <FlorianSchandinat@gmx.de>
Cc: Tomasz Figa <tomasz.figa@gmail.com>
Cc: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agoocfs2: unlock rw lock if inode lock failed
Joseph Qi [Fri, 24 May 2013 22:55:12 +0000 (15:55 -0700)]
ocfs2: unlock rw lock if inode lock failed

In ocfs2_file_aio_write(), it does ocfs2_rw_lock() first and then
ocfs2_inode_lock().

But if ocfs2_inode_lock() failed, it goes to out_sems without unlocking
rw lock.  This will cause a bug in ocfs2_lock_res_free() when testing
res->l_ex_holders, which is increased in __ocfs2_cluster_lock() and
decreased in __ocfs2_cluster_unlock().

Signed-off-by: Joseph Qi <joseph.qi@huawei.com>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Mark Fasheh <mfasheh@suse.com>
Cc: Li Zefan <lizefan@huawei.com>
Cc: "Duyongfeng (B)" <du.duyongfeng@huawei.com>
Acked-by: Sunil Mushran <sunil.mushran@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agomm: mmu_notifier: re-fix freed page still mapped in secondary MMU
Xiao Guangrong [Fri, 24 May 2013 22:55:11 +0000 (15:55 -0700)]
mm: mmu_notifier: re-fix freed page still mapped in secondary MMU

Commit 751efd8610d3 ("mmu_notifier_unregister NULL Pointer deref and
multiple ->release()") breaks the fix 3ad3d901bbcf ("mm: mmu_notifier:
fix freed page still mapped in secondary MMU").

Since hlist_for_each_entry_rcu() is changed now, we can not revert that
patch directly, so this patch reverts the commit and simply fix the bug
spotted by that patch

This bug spotted by commit 751efd8610d3 is:

    There is a race condition between mmu_notifier_unregister() and
    __mmu_notifier_release().

    Assume two tasks, one calling mmu_notifier_unregister() as a result
    of a filp_close() ->flush() callout (task A), and the other calling
    mmu_notifier_release() from an mmput() (task B).

                        A                               B
    t1                                            srcu_read_lock()
    t2            if (!hlist_unhashed())
    t3                                            srcu_read_unlock()
    t4            srcu_read_lock()
    t5                                            hlist_del_init_rcu()
    t6                                            synchronize_srcu()
    t7            srcu_read_unlock()
    t8            hlist_del_rcu()  <--- NULL pointer deref.

This can be fixed by using hlist_del_init_rcu instead of hlist_del_rcu.

The another issue spotted in the commit is "multiple ->release()
callouts", we needn't care it too much because it is really rare (e.g,
can not happen on kvm since mmu-notify is unregistered after
exit_mmap()) and the later call of multiple ->release should be fast
since all the pages have already been released by the first call.
Anyway, this issue should be fixed in a separate patch.

-stable suggestions: Any version that has commit 751efd8610d3 need to be
backported.  I find the oldest version has this commit is 3.0-stable.

[akpm@linux-foundation.org: tweak comments]
Signed-off-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Tested-by: Robin Holt <holt@sgi.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agowait: fix false timeouts when using wait_event_timeout()
Imre Deak [Fri, 24 May 2013 22:55:09 +0000 (15:55 -0700)]
wait: fix false timeouts when using wait_event_timeout()

Many callers of the wait_event_timeout() and
wait_event_interruptible_timeout() expect that the return value will be
positive if the specified condition becomes true before the timeout
elapses.  However, at the moment this isn't guaranteed.  If the wake-up
handler is delayed enough, the time remaining until timeout will be
calculated as 0 - and passed back as a return value - even if the
condition became true before the timeout has passed.

Fix this by returning at least 1 if the condition becomes true.  This
semantic is in line with what wait_for_condition_timeout() does; see
commit bb10ed09 ("sched: fix wait_for_completion_timeout() spurious
failure under heavy load").

Daniel said "We have 3 instances of this bug in drm/i915.  One case even
where we switch between the interruptible and not interruptible
wait_event_timeout variants, foolishly presuming they have the same
semantics.  I very much like this."

One such bug is reported at
  https://bugs.freedesktop.org/show_bug.cgi?id=64133

Signed-off-by: Imre Deak <imre.deak@intel.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: David Howells <dhowells@redhat.com>
Acked-by: Jens Axboe <axboe@kernel.dk>
Cc: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
Cc: Dave Jones <davej@redhat.com>
Cc: Lukas Czerner <lczerner@redhat.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agofat: fix possible overflow for fat_clusters
OGAWA Hirofumi [Fri, 24 May 2013 22:55:08 +0000 (15:55 -0700)]
fat: fix possible overflow for fat_clusters

Intermediate value of fat_clusters can be overflowed on 32bits arch.

Reported-by: Krzysztof Strasburger <strasbur@chkw386.ch.pwr.wroc.pl>
Signed-off-by: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agorapidio: documentation update for enumeration changes
Alexandre Bounine [Fri, 24 May 2013 22:55:07 +0000 (15:55 -0700)]
rapidio: documentation update for enumeration changes

Update RapidIO documentation to reflect changes made to
enumeration/discovery build configuration and user space triggering
mechanism.

Signed-off-by: Alexandre Bounine <alexandre.bounine@idt.com>
Cc: Matt Porter <mporter@kernel.crashing.org>
Cc: Li Yang <leoli@freescale.com>
Cc: Kumar Gala <galak@kernel.crashing.org>
Cc: Andre van Herk <andre.van.herk@Prodrive.nl>
Cc: Micha Nelissen <micha.nelissen@Prodrive.nl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agorapidio: add enumeration/discovery start from user space
Alexandre Bounine [Fri, 24 May 2013 22:55:06 +0000 (15:55 -0700)]
rapidio: add enumeration/discovery start from user space

Add RapidIO enumeration/discovery start from user space.  User space
start allows to defer RapidIO fabric scan until the moment when all
participating endpoints are initialized avoiding mandatory synchronized
start of all endpoints (which may be challenging in systems with large
number of RapidIO endpoints).

Signed-off-by: Alexandre Bounine <alexandre.bounine@idt.com>
Cc: Matt Porter <mporter@kernel.crashing.org>
Cc: Li Yang <leoli@freescale.com>
Cc: Kumar Gala <galak@kernel.crashing.org>
Cc: Andre van Herk <andre.van.herk@Prodrive.nl>
Cc: Micha Nelissen <micha.nelissen@Prodrive.nl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agorapidio: make enumeration/discovery configurable
Alexandre Bounine [Fri, 24 May 2013 22:55:05 +0000 (15:55 -0700)]
rapidio: make enumeration/discovery configurable

Systems that use RapidIO fabric may need to implement their own
enumeration and discovery methods which are better suitable for needs of
a target application.

The following set of patches is intended to simplify process of
introduction of new RapidIO fabric enumeration/discovery methods.

The first patch offers ability to add new RapidIO enumeration/discovery
methods using kernel configuration options.  This new configuration
option mechanism allows to select statically linked or modular
enumeration/discovery method(s) from the list of existing methods or use
external module(s).

This patch also updates the currently existing enumeration/discovery
code to be used as a statically linked or modular method.

The corresponding configuration option is named "Basic
enumeration/discovery" method.  This is the only one configuration
option available today but new methods are expected to be introduced
after adoption of provided patches.

The second patch address a long time complaint of RapidIO subsystem
users regarding fabric enumeration/discovery start sequence.  Existing
implementation offers only a boot-time enumeration/discovery start which
requires synchronized boot of all endpoints in RapidIO network.  While
it works for small closed configurations with limited number of
endpoints, using this approach in systems with large number of endpoints
is quite challenging.

To eliminate requirement for synchronized start the second patch
introduces RapidIO enumeration/discovery start from user space.

For compatibility with the existing RapidIO subsystem implementation,
automatic boot time enumeration/discovery start can be configured in by
specifying "rio-scan.scan=1" command line parameter if statically linked
basic enumeration method is selected.

This patch:

Rework to implement RapidIO enumeration/discovery method selection
combined with ability to use enumeration/discovery as a kernel module.

This patch adds ability to introduce new RapidIO enumeration/discovery
methods using kernel configuration options.  Configuration option
mechanism allows to select statically linked or modular
enumeration/discovery method from the list of existing methods or use
external modules.  If a modular enumeration/discovery is selected each
RapidIO mport device can have its own method attached to it.

The existing enumeration/discovery code was updated to be used as
statically linked or modular method.  This configuration option is named
"Basic enumeration/discovery" method.

Several common routines have been moved from rio-scan.c to make them
available to other enumeration methods and reduce number of exported
symbols.

Signed-off-by: Alexandre Bounine <alexandre.bounine@idt.com>
Cc: Matt Porter <mporter@kernel.crashing.org>
Cc: Li Yang <leoli@freescale.com>
Cc: Kumar Gala <galak@kernel.crashing.org>
Cc: Andre van Herk <andre.van.herk@Prodrive.nl>
Cc: Micha Nelissen <micha.nelissen@Prodrive.nl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agodrivers/block/xsysace.c: fix id with missing port-number
Gernot Vormayr [Fri, 24 May 2013 22:55:03 +0000 (15:55 -0700)]
drivers/block/xsysace.c: fix id with missing port-number

If the port number is missing from the device-tree the device gets named
xs` instead of xsa.  This fixes the check for missing ids.

Tested on ml507 board.

Signed-off-by: Gernot Vormayr <gvormayr@gmail.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Grant Likely <grant.likely@linaro.org>
Cc: Rob Herring <rob.herring@calxeda.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agoMerge tag 'sunxi-fixes-for-3.10' of git://github.com/mripard/linux into fixes
Olof Johansson [Fri, 24 May 2013 20:17:49 +0000 (13:17 -0700)]
Merge tag 'sunxi-fixes-for-3.10' of git://github.com/mripard/linux into fixes

From Maxime Ripard:
Small set of fixes for 3.10:
  - Fix build breakage in pinctrl driver when no other architecture is selected
  - Fix Mini X-plus device tree build

* tag 'sunxi-fixes-for-3.10' of git://github.com/mripard/linux:
  ARM: sunxi: select ARCH_REQUIRE_GPIOLIB
  ARM: sunxi: Fix Mini X-plus device tree build

Signed-off-by: Olof Johansson <olof@lixom.net>
11 years agoxfs: remote attribute lookups require the value length
Dave Chinner [Sun, 19 May 2013 23:51:16 +0000 (09:51 +1000)]
xfs: remote attribute lookups require the value length

When reading a remote attribute, to correctly calculate the length
of the data buffer for CRC enable filesystems, we need to know the
length of the attribute data. We get this information when we look
up the attribute, but we don't store it in the args structure along
with the other remote attr information we get from the lookup. Add
this information to the args structure so we can use it
appropriately.

Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Ben Myers <bpm@sgi.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
(cherry picked from commit e461fcb194172b3f709e0b478d2ac1bdac7ab9a3)

11 years agoxfs: xfs_attr_shortform_allfit() does not handle attr3 format.
Dave Chinner [Sun, 19 May 2013 23:51:14 +0000 (09:51 +1000)]
xfs: xfs_attr_shortform_allfit() does not handle attr3 format.

xfstests generic/117 fails with:

XFS: Assertion failed: leaf->hdr.info.magic == cpu_to_be16(XFS_ATTR_LEAF_MAGIC)

indicating a function that does not handle the attr3 format
correctly. Fix it.

Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Ben Myers <bpm@sgi.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
(cherry picked from commit b38958d715316031fe9ea0cc6c22043072a55f49)

11 years agoxfs: xfs_da3_node_read_verify() doesn't handle XFS_ATTR3_LEAF_MAGIC
Dave Chinner [Sun, 19 May 2013 23:51:13 +0000 (09:51 +1000)]
xfs: xfs_da3_node_read_verify() doesn't handle XFS_ATTR3_LEAF_MAGIC

Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Ben Myers <bpm@sgi.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
(cherry picked from commit 72916fb8cbcf0c2928f56cdc2fbe8c7bf5517758)

11 years agoxfs: fix missing KM_NOFS tags to keep lockdep happy
Dave Chinner [Sun, 19 May 2013 23:51:12 +0000 (09:51 +1000)]
xfs: fix missing KM_NOFS tags to keep lockdep happy

There are several places where we use KM_SLEEP allocation contexts
and use the fact that they are called from transaction context to
add KM_NOFS where appropriate. Unfortunately, there are several
places where the code makes this assumption but can be called from
outside transaction context but with filesystem locks held. These
places need explicit KM_NOFS annotations to avoid lockdep
complaining about reclaim contexts.

Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Ben Myers <bpm@sgi.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
(cherry picked from commit ac14876cf9255175bf3bdad645bf8aa2b8fb2d7c)

11 years agoparisc: fix irq stack on UP and SMP
Helge Deller [Fri, 24 May 2013 21:27:35 +0000 (21:27 +0000)]
parisc: fix irq stack on UP and SMP

The logic to detect if the irq stack was already in use with
raw_spin_trylock() is wrong, because it will generate a "trylock failure
on UP" error message with CONFIG_SMP=n and CONFIG_DEBUG_SPINLOCK=y.

arch_spin_trylock() can't be used either since in the CONFIG_SMP=n case
no atomic protection is given and we are reentrant here. A mutex didn't
worked either and brings more overhead by turning off interrupts.

So, let's use the fastest path for parisc which is the ldcw instruction.

Counting how often the irq stack was used is pretty useless, so just
drop this piece of code.

Signed-off-by: Helge Deller <deller@gmx.de>
11 years agoxfs: Don't reference the EFI after it is freed
Dave Chinner [Sun, 19 May 2013 23:51:10 +0000 (09:51 +1000)]
xfs: Don't reference the EFI after it is freed

Checking the EFI for whether it is being released from recovery
after we've already released the known active reference is a mistake
worthy of a brown paper bag. Fix the (now) obvious use after free
that it can cause.

Reported-by: Dave Jones <davej@redhat.com>
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
(cherry picked from commit 52c24ad39ff02d7bd73c92eb0c926fb44984a41d)

11 years agoxfs: fix rounding in xfs_free_file_space
Dave Chinner [Sun, 19 May 2013 23:51:09 +0000 (09:51 +1000)]
xfs: fix rounding in xfs_free_file_space

The offset passed into xfs_free_file_space() needs to be rounded
down to a certain size, but the rounding mask is built by a 32 bit
variable. Hence the mask will always mask off the upper 32 bits of
the offset and lead to incorrect writeback and invalidation ranges.

This is not actually exposed as a bug because we writeback and
invalidate from the rounded offset to the end of the file, and hence
the offset we are actually punching a hole out of will always be
covered by the code. This needs fixing, however, if we ever want to
use exact ranges for writeback/invalidation here...

Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
(cherry picked from commit 28ca489c63e9aceed8801d2f82d731b3c9aa50f5)

11 years agoxfs: fix sub-page blocksize data integrity writes
Dave Chinner [Sun, 19 May 2013 23:51:08 +0000 (09:51 +1000)]
xfs: fix sub-page blocksize data integrity writes

FSX on 512 byte block size filesystems has been failing for some
time with corrupted data. The fault dates back to the change in
the writeback data integrity algorithm that uses a mark-and-sweep
approach to avoid data writeback livelocks.

Unfortunately, a side effect of this mark-and-sweep approach is that
each page will only be written once for a data integrity sync, and
there is a condition in writeback in XFS where a page may require
two writeback attempts to be fully written. As a result of the high
level change, we now only get a partial page writeback during the
integrity sync because the first pass through writeback clears the
mark left on the page index to tell writeback that the page needs
writeback....

The cause is writing a partial page in the clustering code. This can
happen when a mapping boundary falls in the middle of a page - we
end up writing back the first part of the page that the mapping
covers, but then never revisit the page to have the remainder mapped
and written.

The fix is simple - if the mapping boundary falls inside a page,
then simple abort clustering without touching the page. This means
that the next ->writepage entry that write_cache_pages() will make
is the page we aborted on, and xfs_vm_writepage() will map all
sections of the page correctly. This behaviour is also optimal for
non-data integrity writes, as it results in contiguous sequential
writeback of the file rather than missing small holes and having to
write them a "random" writes in a future pass.

With this fix, all the fsx tests in xfstests now pass on a 512 byte
block size filesystem on a 4k page machine.

Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
(cherry picked from commit 49b137cbbcc836ef231866c137d24f42c42bb483)

11 years agoparisc/superio: Use module_pci_driver to register driver
Peter Huewe [Mon, 20 May 2013 20:56:45 +0000 (20:56 +0000)]
parisc/superio: Use module_pci_driver to register driver

Removing some boilerplate by using module_pci_driver instead of calling
register and unregister in the otherwise empty init/exit functions.

Signed-off-by: Peter Huewe <peterhuewe@gmx.de>
Signed-off-by: Helge Deller <deller@gmx.de>
11 years agoparisc: make interrupt and interruption stack allocation reentrant
John David Anglin [Mon, 20 May 2013 16:42:53 +0000 (16:42 +0000)]
parisc: make interrupt and interruption stack allocation reentrant

The get_stack_use_cr30 and get_stack_use_r30 macros allocate a stack
frame for external interrupts and interruptions requiring a stack frame.
They are currently not reentrant in that they save register context
before the stack is set or adjusted.

I have observed a number of system crashes where there was clear
evidence of stack corruption during interrupt processing, and as a
result register corruption. Some interruptions can still occur during
interruption processing, however external interrupts are disabled and
data TLB misses don't occur for absolute accesses. So, it's not entirely
clear what triggers this issue. Also, if an interruption occurs when
Q=0, it is generally not possible to recover as the shadowed registers
are not copied.

The attached patch reworks the get_stack_use_cr30 and get_stack_use_r30
macros to allocate stack before doing register saves. The new code is a
couple of instructions shorter than the old implementation. Thus, it's
an improvement even if it doesn't fully resolve the stack corruption
issue. Based on limited testing, it improves SMP system stability.

Signed-off-by: John David Anglin <dave.anglin@bell.net>
Signed-off-by: Helge Deller <deller@gmx.de>
11 years agoparisc: show number of FPE and unaligned access handler calls in /proc/interrupts
Helge Deller [Sat, 18 May 2013 19:35:44 +0000 (19:35 +0000)]
parisc: show number of FPE and unaligned access handler calls in /proc/interrupts

Show number of floating point assistant and unaligned access fixup
handler in /proc/interrupts file.

Signed-off-by: Helge Deller <deller@gmx.de>
11 years agoparisc: add additional parisc git tree to MAINTAINERS file
Helge Deller [Thu, 16 May 2013 21:00:16 +0000 (21:00 +0000)]
parisc: add additional parisc git tree to MAINTAINERS file

Signed-off-by: Helge Deller <deller@gmx.de>
11 years agoparisc: use PAGE_SHIFT instead of hardcoded value 12 in pacache.S
Helge Deller [Thu, 16 May 2013 20:51:41 +0000 (20:51 +0000)]
parisc: use PAGE_SHIFT instead of hardcoded value 12 in pacache.S

additionally clean up some whitespaces & tabs.

Signed-off-by: Helge Deller <deller@gmx.de>
11 years agoparisc: add rp5470 entry to machine database
Helge Deller [Thu, 16 May 2013 20:42:39 +0000 (20:42 +0000)]
parisc: add rp5470 entry to machine database

Signed-off-by: Helge Deller <deller@gmx.de>
11 years agoMPILIB: disable usage of floating point registers on parisc
Helge Deller [Sun, 5 May 2013 22:08:22 +0000 (22:08 +0000)]
MPILIB: disable usage of floating point registers on parisc

The umul_ppmm() macro for parisc uses the xmpyu assembler statement
which does calculation via a floating point register.

But usage of floating point registers inside the Linux kernel are not
allowed and gcc will stop compilation due to the -mdisable-fpregs
compiler option.

Fix this by disabling the umul_ppmm() and udiv_qrnnd() macros. The
mpilib will then use the generic built-in implementations instead.

Signed-off-by: Helge Deller <deller@gmx.de>
11 years agoMerge tag 'trace-fixes-v3.10-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Fri, 24 May 2013 17:46:55 +0000 (10:46 -0700)]
Merge tag 'trace-fixes-v3.10-rc2' of git://git./linux/kernel/git/rostedt/linux-trace

Pull tracing fix from Steven Rostedt:
 "Masami Hiramatsu fixed another bug.  This time returning a proper
  result in event_enable_func().  After checking the return status of
  try_module_get(), it returned the status of try_module_get().

  But try_module_get() returns 0 on failure, which is success for
  event_enable_func()"

* tag 'trace-fixes-v3.10-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace:
  tracing: Return -EBUSY when event_enable_func() fails to get module

11 years agoMerge branch 'for-next' of git://git.samba.org/sfrench/cifs-2.6
Linus Torvalds [Fri, 24 May 2013 17:45:59 +0000 (10:45 -0700)]
Merge branch 'for-next' of git://git.samba.org/sfrench/cifs-2.6

Pull CIFS fix from Steve French:
 "One cifs fix to merge now - fixes possible DFS oops (I expect to
  request a merge of 4 additional cifs fixes next week)"

* 'for-next' of git://git.samba.org/sfrench/cifs-2.6:
  cifs: only set ops for inodes in I_NEW state

11 years agoMerge tag 'at91-fixes' of git://github.com/at91linux/linux-at91 into fixes
Olof Johansson [Fri, 24 May 2013 17:11:44 +0000 (10:11 -0700)]
Merge tag 'at91-fixes' of git://github.com/at91linux/linux-at91 into fixes

From Nicolas Ferre:
- One definition fix that can lead to mis-clock some AT91 peripherals on SAMA5.
- Two DT related fixes.

* tag 'at91-fixes' of git://github.com/at91linux/linux-at91:
  ARM: at91/sama5: fix incorrect PMC pcr div definition
  ARM: at91/dt: fix macb pinctrl_macb_rmii_mii_alt definition
  ARM: at91: at91sam9n12: move external irq declatation to DT

Signed-off-by: Olof Johansson <olof@lixom.net>